From: Yen Chi Hsuan Date: Fri, 26 May 2017 06:15:46 +0000 (+0800) Subject: [youku] Fix extraction (closes #13191) X-Git-Url: https://git.cielonegro.org/gitweb.cgi?a=commitdiff_plain;h=59ed87cbd9ea08c889514a05b646141004f432a1;p=youtube-dl.git [youku] Fix extraction (closes #13191) --- diff --git a/ChangeLog b/ChangeLog index 9e4f53093..6a05657ab 100644 --- a/ChangeLog +++ b/ChangeLog @@ -3,6 +3,9 @@ version Core * [postprocessor/ffmpeg] Fix metadata filename handling on Python 2 (#13182) +Extractors +* [youku] Fix extraction (#13191) + version 2017.05.23 diff --git a/youtube_dl/extractor/youku.py b/youtube_dl/extractor/youku.py index 73ebe5759..37be440fa 100644 --- a/youtube_dl/extractor/youku.py +++ b/youtube_dl/extractor/youku.py @@ -1,7 +1,6 @@ # coding: utf-8 from __future__ import unicode_literals -import base64 import itertools import random import re @@ -9,15 +8,9 @@ import string import time from .common import InfoExtractor -from ..compat import ( - compat_ord, - compat_str, - compat_urllib_parse_urlencode, -) from ..utils import ( ExtractorError, get_element_by_attribute, - try_get, ) @@ -35,9 +28,9 @@ class YoukuIE(InfoExtractor): # MD5 is unstable 'url': 'http://v.youku.com/v_show/id_XMTc1ODE5Njcy.html', 'info_dict': { - 'id': 'XMTc1ODE5Njcy_part1', + 'id': 'XMTc1ODE5Njcy', 'title': '★Smile﹗♡ Git Fresh -Booty Music舞蹈.', - 'ext': 'flv' + 'ext': 'mp4', } }, { 'url': 'http://player.youku.com/player.php/sid/XNDgyMDQ2NTQw/v.swf', @@ -46,25 +39,24 @@ class YoukuIE(InfoExtractor): 'url': 'http://v.youku.com/v_show/id_XODgxNjg1Mzk2_ev_1.html', 'info_dict': { 'id': 'XODgxNjg1Mzk2', + 'ext': 'mp4', 'title': '武媚娘传奇 85', }, - 'playlist_count': 11, - 'skip': 'Available in China only', }, { 'url': 'http://v.youku.com/v_show/id_XMTI1OTczNDM5Mg==.html', 'info_dict': { 'id': 'XMTI1OTczNDM5Mg', + 'ext': 'mp4', 'title': '花千骨 04', }, - 'playlist_count': 13, }, { 'url': 'http://v.youku.com/v_show/id_XNjA1NzA2Njgw.html', 'note': 'Video protected with password', 'info_dict': { 'id': 'XNjA1NzA2Njgw', + 'ext': 'mp4', 'title': '邢義田复旦讲座之想象中的胡人—从“左衽孔子”说起', }, - 'playlist_count': 19, 'params': { 'videopassword': '100600', }, @@ -73,130 +65,16 @@ class YoukuIE(InfoExtractor): 'url': 'http://v.youku.com/v_show/id_XOTUxMzg4NDMy.html', 'info_dict': { 'id': 'XOTUxMzg4NDMy', + 'ext': 'mp4', 'title': '我的世界☆明月庄主☆车震猎杀☆杀人艺术Minecraft', }, - 'playlist_count': 6, }] - def construct_video_urls(self, data): - # get sid, token - def yk_t(s1, s2): - ls = list(range(256)) - t = 0 - for i in range(256): - t = (t + ls[i] + compat_ord(s1[i % len(s1)])) % 256 - ls[i], ls[t] = ls[t], ls[i] - s = bytearray() - x, y = 0, 0 - for i in range(len(s2)): - y = (y + 1) % 256 - x = (x + ls[y]) % 256 - ls[x], ls[y] = ls[y], ls[x] - s.append(compat_ord(s2[i]) ^ ls[(ls[x] + ls[y]) % 256]) - return bytes(s) - - sid, token = yk_t( - b'becaf9be', base64.b64decode(data['security']['encrypt_string'].encode('ascii')) - ).decode('ascii').split('_') - - # get oip - oip = data['security']['ip'] - - fileid_dict = {} - for stream in data['stream']: - if stream.get('channel_type') == 'tail': - continue - format = stream.get('stream_type') - fileid = try_get( - stream, lambda x: x['segs'][0]['fileid'], - compat_str) or stream['stream_fileid'] - fileid_dict[format] = fileid - - def get_fileid(format, n): - number = hex(int(str(n), 10))[2:].upper() - if len(number) == 1: - number = '0' + number - streamfileids = fileid_dict[format] - fileid = streamfileids[0:8] + number + streamfileids[10:] - return fileid - - # get ep - def generate_ep(format, n): - fileid = get_fileid(format, n) - ep_t = yk_t( - b'bf7e5f01', - ('%s_%s_%s' % (sid, fileid, token)).encode('ascii') - ) - ep = base64.b64encode(ep_t).decode('ascii') - return ep - - # generate video_urls - video_urls_dict = {} - for stream in data['stream']: - if stream.get('channel_type') == 'tail': - continue - format = stream.get('stream_type') - video_urls = [] - for dt in stream['segs']: - n = str(stream['segs'].index(dt)) - param = { - 'K': dt['key'], - 'hd': self.get_hd(format), - 'myp': 0, - 'ypp': 0, - 'ctype': 12, - 'ev': 1, - 'token': token, - 'oip': oip, - 'ep': generate_ep(format, n) - } - video_url = \ - 'http://k.youku.com/player/getFlvPath/' + \ - 'sid/' + sid + \ - '_00' + \ - '/st/' + self.parse_ext_l(format) + \ - '/fileid/' + get_fileid(format, n) + '?' + \ - compat_urllib_parse_urlencode(param) - video_urls.append(video_url) - video_urls_dict[format] = video_urls - - return video_urls_dict - @staticmethod def get_ysuid(): return '%d%s' % (int(time.time()), ''.join([ random.choice(string.ascii_letters) for i in range(3)])) - def get_hd(self, fm): - hd_id_dict = { - '3gp': '0', - '3gphd': '1', - 'flv': '0', - 'flvhd': '0', - 'mp4': '1', - 'mp4hd': '1', - 'mp4hd2': '1', - 'mp4hd3': '1', - 'hd2': '2', - 'hd3': '3', - } - return hd_id_dict[fm] - - def parse_ext_l(self, fm): - ext_dict = { - '3gp': 'flv', - '3gphd': 'mp4', - 'flv': 'flv', - 'flvhd': 'flv', - 'mp4': 'mp4', - 'mp4hd': 'mp4', - 'mp4hd2': 'flv', - 'mp4hd3': 'flv', - 'hd2': 'flv', - 'hd3': 'flv', - } - return ext_dict[fm] - def get_format_name(self, fm): _dict = { '3gp': 'h6', @@ -210,32 +88,40 @@ class YoukuIE(InfoExtractor): 'hd2': 'h2', 'hd3': 'h1', } - return _dict[fm] + return _dict.get(fm) def _real_extract(self, url): video_id = self._match_id(url) self._set_cookie('youku.com', '__ysuid', self.get_ysuid()) + self._set_cookie('youku.com', 'xreferrer', 'http://www.youku.com') - def retrieve_data(req_url, note): - headers = { - 'Referer': req_url, - } - headers.update(self.geo_verification_headers()) - self._set_cookie('youku.com', 'xreferrer', 'http://www.youku.com') - - raw_data = self._download_json(req_url, video_id, note=note, headers=headers) + _, urlh = self._download_webpage_handle( + 'https://log.mmstat.com/eg.js', video_id, 'Retrieving cna info') + # The etag header is '"foobar"'; let's remove the double quotes + cna = urlh.headers['etag'][1:-1] - return raw_data['data'] + # request basic data + basic_data_params = { + 'vid': video_id, + 'ccode': '0401', + 'client_ip': '192.168.1.1', + 'utid': cna, + 'client_ts': time.time() / 1000, + } video_password = self._downloader.params.get('videopassword') - - # request basic data - basic_data_url = 'http://play.youku.com/play/get.json?vid=%s&ct=12' % video_id if video_password: - basic_data_url += '&pwd=%s' % video_password + basic_data_params['password'] = video_password - data = retrieve_data(basic_data_url, 'Downloading JSON metadata') + headers = { + 'Referer': url, + } + headers.update(self.geo_verification_headers()) + data = self._download_json( + 'https://ups.youku.com/ups/get.json', video_id, + 'Downloading JSON metadata', + query=basic_data_params, headers=headers)['data'] error = data.get('error') if error: @@ -255,37 +141,21 @@ class YoukuIE(InfoExtractor): # get video title title = data['video']['title'] - # generate video_urls_dict - video_urls_dict = self.construct_video_urls(data) - - # construct info - entries = [{ - 'id': '%s_part%d' % (video_id, i + 1), - 'title': title, - 'formats': [], - # some formats are not available for all parts, we have to detect - # which one has all - } for i in range(max(len(v.get('segs')) for v in data['stream']))] - for stream in data['stream']: - if stream.get('channel_type') == 'tail': - continue - fm = stream.get('stream_type') - video_urls = video_urls_dict[fm] - for video_url, seg, entry in zip(video_urls, stream['segs'], entries): - entry['formats'].append({ - 'url': video_url, - 'format_id': self.get_format_name(fm), - 'ext': self.parse_ext_l(fm), - 'filesize': int(seg['size']), - 'width': stream.get('width'), - 'height': stream.get('height'), - }) + formats = [{ + 'url': stream['m3u8_url'], + 'format_id': self.get_format_name(stream.get('stream_type')), + 'ext': 'mp4', + 'protocol': 'm3u8_native', + 'filesize': int(stream.get('size')), + 'width': stream.get('width'), + 'height': stream.get('height'), + } for stream in data['stream'] if stream.get('channel_type') != 'tail'] + self._sort_formats(formats) return { - '_type': 'multi_video', 'id': video_id, 'title': title, - 'entries': entries, + 'formats': formats, }